Events Calendar

SE PhD Final Defense: Jonas Hall

TITLE: Decomposing Persistent Monitoring Problems using Numerical Optimal Control

ADVISOR: Sean Andersson ME, SE

COMMITTEE: Roberto Tron ME, SE; Christos Cassandras ECE; Roberto Tron ME; Alexander Olshevsky ECE; Alyssa Pierson ME

ABSTRACT: This thesis addresses the challenge of optimizing Persistent Monitoring (PM) missions by decomposing complex control problems using numerical optimal control techniques. Persistent monitoring involves continuously gathering information about dynamically evolving systems, with applications in environmental monitoring, disaster response, surveillance, and more. Persistent monitoring missions inherently involve challenging trade-offs between exploration and exploitation, as agents must balance visiting multiple locations against thoroughly observing dynamic targets. This complexity is compounded by the requirement to optimize both discrete decisions, such as determining the order of target visits, and continuous decisions related to agent trajectories. Such considerations render the direct solution of PM problems difficult, particularly in multi-dimensional settings. A hierarchical decomposition that leverages a typical structure in PM problems is proposed to efficiently address these challenges. First, foundational theoretical results in one-dimensional PM scenarios are established, demonstrating that optimal controls can be represented via finite-dimensional parameterizations. Building upon this, the approach is generalized to two-dimensional mission spaces by casting PM problems into bilevel optimization problems, wherein lower-level subproblems are simplified optimal control problems, and upper-level problems optimize trajectory boundary conditions. This decomposition significantly reduces computational complexity, enabling real-time applicability and adaptability.